# HeLaCytoNuc: fluorescence microscopy of HeLa cells with nucleus and whole-cell segmentation masks (RODARE)

## Facts

- Provider: Trina De, Artur Yakimovich et al. (HZDR / CASUS); packaged by SPORA (csmp-hub)
- License: CC-BY-4.0 (https://rodare.hzdr.de/record/3001)
- Size: 15 files
- Format: PNG, JSON, Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/csmp-hub/helacytonuke
- Access: public, readable without an account
- DOI: https://doi.org/10.14278/rodare.3001
- Category: Life Sciences
- Source: https://doi.org/10.14278/rodare.3001
- Page: https://scigantic.com/archive/0dace17f-98bf-5ff2-a708-bd945f08503e
- Keywords: Life Sciences, image segmentation, PNG, JSON, Parquet, Hugging Face

## About

HeLaCytoNuc is a fluorescence microscopy dataset of cultured HeLa cells with instance masks for both nuclei and whole cells, released by the Helmholtz-Zentrum Dresden-Rossendorf group of Artur Yakimovich. It has 2,676 images of 520x696 pixels in which nuclei appear blue and actin fibers red; the images come from a large-scale RNAi screen for pathogen entry (Ramo et al. 2014). Training and validation masks were generated with CellProfiler, and test masks were drawn by hand, with the CellProfiler masks for the same test images kept alongside for comparison. Each row stores the image and 16-bit nucleus and whole-cell instance masks as PNG bytes, a binary nucleus mask, and 114,733 computed nucleus-to-cell pairings. The original train, validation and test splits are kept in train.parquet, val.parquet and test.parquet, about 710 MB (measured 2026-11-21); read with pandas or polars plus Pillow, or Hugging Face datasets. Use for nucleus and cytoplasm segmentation, cell instance segmentation and testing models trained on automatic labels against manual ones. Repackaged by the SPORA project (csmp-hub) from RODARE record 3001, released under CC BY 4.0 with DOI 10.14278/rodare.3001; cite De et al. 2024.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("csmp-hub/helacytonuke")
print(ds)
```

## Data structure

File types in sampled files: PNG, JSON, Parquet

Example file: `test.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/0dace17f-98bf-5ff2-a708-bd945f08503e
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/0dace17f-98bf-5ff2-a708-bd945f08503e
- Over MCP: `get_schema_card` with id `0dace17f-98bf-5ff2-a708-bd945f08503e` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [JSON datasets](https://scigantic.com/topics/json)
- [Image segmentation datasets](https://scigantic.com/topics/image-segmentation)
- [PNG datasets](https://scigantic.com/topics/png)
- [Life Sciences datasets](https://scigantic.com/topics/life-sciences)
